Nvidia data center chief signals acceleration in on-premises GPU deployments for AI applications.

Ian Buck, VP and general manager of Nvidia's data center business, discussed the evolving landscape of AI infrastructure deployment with Data Center Knowledge. Buck's comments come as Nvidia's data center business reported 71 percent more revenue than last year, reflecting strong demand for GPU infrastructure across industries.

According to Buck, enterprises are increasingly deploying GPUs on-premises rather than relying exclusively on cloud providers. "Certainly. We're seeing folks like SAP, Salesforce, eBay start to train neural networks and deploy their neural networks on GPUs for inferencing in on-prem environments," Buck said. The decision to maintain on-premises infrastructure stems from multiple factors. "Some IT departments are quite capable of managing their own infrastructure. They can optimize it and customize it specifically to their own needs, not just what the CSPs offer," Buck explained. Data privacy concerns also drive the choice: "They also have concerns about data privacy, especially financial information, user information. It's more comforting to have data on their own premises, on their own servers that they control." Buck noted that "It's a matter of preference and choice, and sometimes economics as well. Certainly, it's a rent-versus-buy decision, and the economics is different in every case."

Different industries are adopting on-premises GPU infrastructure at varying rates. Self-driving vehicles represent "a large growth area," according to Buck, as companies need "large systems for simulating and training these neural networks." Healthcare has emerged as another major sector, with "a massive amount of medical data" driving investment in deep learning applications for disease understanding, cancer research, diagnosis, and medical imaging. In traditional data analytics, companies like SAP are "investing heavily in providing new kinds of services like brand insight, where they can actually recognize brands in video streams and sporting events," while also deploying deep neural networks for "recommender systems, for better advertising, for better click rates."

Buck identified hyperscalers, SAP, and GE as the most advanced practitioners. GE, in particular, is "doing really well" with predictive maintenance, using AI to "ingest all of the data that's coming off sensors on gas turbines used to generate energy" and service equipment before failure. The company is also "doing some interesting work with drones to observe and look for failures on sites like oil rigs to identify problems early and take action before something more expensive or more catastrophic could occur."

Looking forward, Buck expects a persistent mix of on-premises and cloud infrastructure. "No, I think there will always be a mix. I'm not sure what the actual mix will end up being," he said. Some enterprises view AI infrastructure ownership as strategically important, allowing them to "tune and design their data center architecture, the infrastructure, the InfiniBand switches, and the network topologies to build perhaps more supercomputers than straight-up clusters." The fundamental drivers of on-premises deployment remain unchanged: IT capability, architectural optimization opportunities that cloud providers don't offer, data privacy requirements, and economics of long-term asset ownership versus rental.